Wednesday, January 28, 2026Major ReleaseHigh Impact
GeminiVersion 3 Ultra

Google Launches Gemini 3 Ultra with Native Multimodal Reasoning

Google DeepMind released Gemini 3 Ultra, its most capable model featuring native multimodal reasoning across text, image, audio, and video. The model leads on vision benchmarks and introduces real-time video understanding.

Key Highlights

1

Native multimodal reasoning across text, image, audio, video

2

Real-time video understanding in Google AI Studio

3

95.2% on DocVQA (new state of the art)

4

2M token context window

← Back to changelog